The Transmembrane Domain of Glycophorin A as Studied by Cross-linking Using Photoactivatable Phospholipids*

Glycophorin A, the major sialoglycoprotein of the human erythrocyte, consists of a NHz-terminal carbo-hydrate-rich region exposed to the outside, a hydro- phobic region which forms a transmembrane bridge, and a COOH-terminal hydrophilic region extending into the cytoplasm. With the aim of further defining the membrane-embedded region, the protein has been re- constituted into vesicles formed from dimyristoylphos-phatidylcholine and phospholipids containing pho- tosensitive carbene precursors. The photosensitive groups were incorporated either at the u-position of sn-2 fatty acyl chain or the polar head group of leci-thins. Following photolysis, covalent cross-linking (1-2%) of the photoactivatable phospholipids to the protein was demonstrated. Degradation and sequence analysis showed that in the case of phospholipids containing photoactivatable groups in the fatty acyl chains most of the covalent cross-linking involved the carboxyl group of Glu-70. Therefore, the latter residue must be within the bilayer. This conclusion was supported by the reaction of the membrane-permeant [‘“C] dicyclohexylcarbodiimide with glycophorin reconsti- tuted into vesicles. The same residue was labeled. Photolysis of glycophorin vesicles containing phospholip- ids with photolabels in the polar head group gave products in which the cross-links

into the cytoplasm. With the aim of further defining the membrane-embedded region, the protein has been reconstituted into vesicles formed from dimyristoylphosphatidylcholine and phospholipids containing photosensitive carbene precursors. The photosensitive groups were incorporated either at the u-position of sn-2 fatty acyl chain or the polar head group of lecithins. Following photolysis, covalent cross-linking (1-2%) of the photoactivatable phospholipids to the protein was demonstrated. Degradation and sequence analysis showed that in the case of phospholipids containing photoactivatable groups in the fatty acyl chains most of the covalent cross-linking involved the carboxyl group of Glu-70. Therefore, the latter residue must be within the bilayer. This conclusion was supported by the reaction of the membrane-permeant ['"C] dicyclohexylcarbodiimide with glycophorin reconstituted into vesicles. The same residue was labeled. Photolysis of glycophorin vesicles containing phospholipids with photolabels in the polar head group gave products in which the cross-links were present in peptide fragments (residues 62-81 and 82-96). These results define the probable boundaries of the membraneembedded segment of glycophorin A. Corresponding experiments with erythrocyte ghosts gave similar results. Glycophorin A, the major sialoglycoprotein of the human erythrocytes, is a widely studied and well characterized membrane protein (1). The protein, which consists of a single polypeptide chain of 131 amino acids, has been shown to have a tripartite structure (2). The NH2 terminus of this protein is exposed to the outside of the erythrocyte membrane and carries all of the carbohydrate (3). The protein then traverses the membrane through a predominantly hydrophobic region of about 20 amino acids and, finally, there is a hydrophilic segment at the COOH terminus which is on the cytoplasmic face of the membrane (4). Further, the total amino acid sequence of the protein has been determined (2). Because of this chemical progress, glycophorin A is especially attractive for further structural studies.
A general chemical approach that aims at covalent crosslinking between membrane proteins and phospholipids is being investigated in this laboratory (5). Photoactivatable groups capable of generating reactive earbene intermediates are incorporated into phospholipids as structural components in two types of phospholipids. In one type, the photoactivatable groups are introduced by chemical synthesis (6, 7) into the w-position of the fatty acid present at the sn-2 position of the glycerol backbone (phospholipids . The aim in the use of these phospholipids is to achieve covalent cross-linking between the intramembranous regions of the membrane proteins and phospholipids (8,9). In the second type of photoactivatable phospholipids, the label is introduced in the polar head groups of the phospholipids (e.g. phospholipids IV and V) (7). Proteins reconstituted into vesicles containing such phospholipids should form cross-links with segments of the polypeptide chains immediately exposed to the outside of the membrane.
In this paper, we report on the use of the photoactivatable phospholipids (I-V, Fig. 1) in a study of the membrane-embedded segment and its flanking regions in glycophorin A. The results using phospholipids  showed, in particular, that Glu-70 is embedded in the membrane and this conclusion was further supported by the reaction of [14C]dicyclohexylcarbodiimide with glycophorin reconstituted into vesicles. The use of phospholipids IV and V ( Fig. 1) provided additional insight into the regions flanking the membrane in the reconstituted glycophorin vesicles. Based on these studies and on previous physical and biochemical data, a model is proposed for the transmembrane structure of glycophorin A. The model places residues 63-94 within the membrane in an a-helical conformation. Several polar residues which are embedded within the bilayer may form salt bridges or hydrogen bonds to mask their hydrophilicity.

The Transmembrane Domain of Glycophorin A
macia h e Chemicals. Bio-Gel A-1.5m, Bio-Gel A-15m, and Bio-Beads SM-2 were from Bio-Rad Corp. Wheat germ agglutinin-Sephm w was obtained from Vector Laboratories. Plastic-backed silica gel plates (Silica Gel 60) was purchased from E. M. Laboratories.
Dodecyldimethylammoniium oxide was purchased from Onyx Chemical Co. as a 30% aqueous solution under the name Ammonyx LO. The detergent was extracted with organic solvents (10) and was then repeatedly dried from ethanol to remove traces of water. The detergent was further p u d e d as described (11).
Synthesis of Phospholipids, Fatty Acids, and [14C]Dicyclohexylcarbodiimide-Phospholipids containing m-diazirinophenoxy groups in the fatty acyl chain and pyridyl diazirine in the polar head group were synthesized as described (7). ["CIDMPC labeled in the N"CH3 groups was synthesized by published procedures (12). Specific activities of phospholipid I1 (Fig. 1) and ["CIDMPC were 48 and 4.4 mCi/ mmol, respectively. The specific activities of phospholipids I, 111, IV, and V ( Fig. 1) were 5.6 mCi/mmol. Fatty acid derivative VI1 was synthesized by coupling m-hydroxybenzaldehyde with ethyl w-iodoundecanoate followed by sodium borohydride reduction. Alkaline hydrolysis of Compound VI1 afforded the fatty acid VIII. ["CIDCC (55 mCi/mol) was synthesized according to published procedures (13).
Isolation of Glycophorin A-Human blood (not outdated) was obtained from the Boston Red Cross. A mixture of glycophorin A, B, and C from the erythrocytes was prepared according to Marchesi and Andrews (14). Glycophorin A was purified by passage through a column (2.6 X 94 cm) of Bio-Gel A-1.5m equilibrated with 0.1% (w/v) dodecyldimethylammonium oxide, 25 m~ NaCl, 5 n" sodium phosphate, pH 8.0,0.01% NaN3 (15). Finally, the glycophorin A preparation was delipidated by successive extractions with chloroform/methanol (Ll, v/v) and chloroform/methanol/concentrated HC1 (1501501, v/ v/v).
Amino acid analysis of glycophorin A prepared as above was consistent with the composition calculated from the amino acid sequence previously reported (2). The sialic acid content of the preparation was 21% (w/w), and the phosphate content was 0.42 f 0.05 mol/mol. Glycophorin A showed a weak ultraviolet absorbance (EZX 3.4 in 100 m~ TI&, pH 7.5).
Incorporation of Glycophorin A into Artificial Vesicles-Single bilayer vesicles containing glycophorin A and phospholipids were prepared by the cholate dialysis method (ls).' In a typical experiment, 1.0 mg of DMPC or a mixture of DMPC and radioactive diazirinecontaining phospholipid (e.g. 111, Fig. 1) (usually at a ratio of 9 1 by weight) in chloroform solution was dried under nitrogen and the residue was suspended in 1.25 ml of 10 m~ sodium cholate, pH 7.5, by vigorous Vortex mixing. The solution was sonicated for about 15 min in a bath sonicator (Laboratory Supplies Co., Hicksville, NY). A solution of glycophorin A (0.2 mg in 0.1 ml of water) was mixed with 10 mM sodium cholate (0.1 ml), pH 7.5. The sonicated lipid solution was then mixed with the protein solution; the solution was kept at against 1 liter of 100 n" Tris, pH 7.5, containing 10 ml of SM-2 Bio-room temperature for 1 h and then dialyzed at room temperature Beads that had been extensively washed with water. Following a change of the buffer after 24 h, the dialysis was continued for another 24 h. The vesicles were collected by centrifugation ( 1 0 0 , O O O X g) for 0.5 h, resuspended in the same buffer (0.25 ml), and stored at 4 "C.
Analysis of Residual Cholic Acid in Vesicles-The amount of ' S. Kirchanski, unpublished work.
cholic acid that remainedin the dialyzed vesicle preparation following the above dialysis procedure was monitored by including some [3H] cholic acid. Following extraction of the lipids (IO), the cholic acid was determined by thin layer chromatography (Solvent 1) and fluorogra-Sucrose Density Gradient Centrifugation-Samples of vesicles containing glycophorin A, DMPC, and a diazirine-carrying phospholipid (Fig. 1) were layered on top of a linear 6-cm sucrose density gradient (5-30% w/w). They were centrifuged at 149,000 X g for 15-24 h. The phospholipid distribution was measured by including either [I4C]DMPC or radioactive phospholipid I (Fig. 1) during the preparation of the vesicles. Glycophorin A was measured by analysis of sialic acid after removal of sucrose by dialysis.
Orientation of Glycophorin A in Reconstituted Vesicles-Vesicles containing 12 pg of glycophorin A were incubated at room temperature with 0.019 unit of C. perfringens neuraminidase. The release of sialic acid was measured by one of two methods. In the first method, the reaction was terminated by adding SDS to 1% (w/v) concentration and sialic acid was assayed directly (17). No release of sialic acid was observed when neuraminidase was added to vesicles predissolved in 1% SDS. In the second method, 1.9 ml of 1 0 0 m~ triethylamine bicarbonate (pH P) at 0 "C were added to the reaction mixture. The vesicles were pelleted by centrifugation (105,000 X g for 30 mid, and after lyophilization, the sialic acid contents of the pellet and of the supernatant solution were determined. Papain was also used to determine the orientation of glycophorin A in the vesicles. After activation with 0.5 m~ EDTA and 0.5 m~ cysteine, mercuripapain was .jncubated at room temperature with the vesicles at a protease/glycophorin A ratio of 1:40 (w/w). The reaction was terminated by the addition of 70 m~ iodoacetic acid, and the sample was centrifuged as described above. Sialic acid was assayed after hydrolysis with 0.1 N H2S04 at 80 "C for 1 h.
Photolyses-Light from a 1000 W arc lamp from Oriel Corp. was passed successively through an fD.7 primary focusing lens, a water heat filter, an f2.0 secondary focusing lens, a monochromator set at 366 nm, and a 345 nm cut-off filter from Schott Optical Glass, Inc. Unless noted otherwise, the sample was irradiated in a quartz cuvette held in a thermostated metal block. Water-saturated nitrogen was blown into the cuvette for 20 min prior to and during the photolysis.
Under these conditions, the ultraviolet spectra of both bovine serum albumin and free tryptophan were unchanged after 10 min of photoly s i s . Unless stated otherwise, samples containing phospholipids I, 11, and I11 ( Fig. 1) were irradiated at 30 "C for 4 min, and samples containing phospholipids IV and V for 10 min. The length of time between the termination of photolysis and solubilization of the vesicles (organic solvents) or detergents was at least 30 min unless stated otherwise.
Separation of Cross-linked Glycophorin A from Free Phospholipids-After photolysis, the solution containing the vesicles was lyophilized and the residue was dissolved in 88% formic acid/ethanol/water (20: 50:30, v/v/v) or in some cases 88% formic acid and ethanol were added directly to form a solution with the same final concentration. The cross-linked glycophorin A was separated from the phospholipids on a Sephadex LH-60 column (1.0 X 45 cm) equilibrated with the same solvent. Fractions of 0.8 ml were collected every 5 min.
Reaction of ["C]DCC with Glycophorin A in Reconstituted Vesicles-Vesicles containing 0.21 nmol of glycophorin A and 46 nmol of DMPC were suspended in 0.2 ml of 100 m~ Tris, pH 7.5. ["C]DCC (2-78 nmol) dissolved in ethanol was added to the above suspension and then kept at room temperature for 16-18 h. The final ethanol content was 1% or less. Then formic acid was added to 22% (v/v) and the total solution was applied to a Sephadex LH-60 column (0.7 X 18 cm) equilibrated with 88% formic acid/ethanol/water (20W30, v/v/ V) .
Fragmentation of Glycophorin A-CNBr cleavages of small amounts (50 pg) of cross-linked glycophorin A were carried out in 25 pl of a stock solution of CNBr in 70% formic acid (25 mg/ml) for 48 h at room temperature. out for 16 h at 37 "C in 1 0 0 m~ Tris, pH 7. 5 (18). The weight ratio of Digestion of the cross-linked glycophorin A with trypsin was carried the enzyme to the cross-linked protein was 1:30. The solution was adjusted to pH 4 with acetic acid and then filtered through a 0.45-p Millipore filter. After rinsing the Nter three times with 0.1 m~ HCl, an aliquot of the filtrate was counted for radioactivity. Proteolysis of Glycophorin A in Photolyzed Vesicles and the Isolation of Cross-linked Peptides-Mercuripapain, which had been activated with 0.5 m~ EDTA, 0.5 m~ cysteine, was added to photolyzed glycophorin A-containing vesicles at a ratio of 1 mg of papain/ phy. by guest on March 23, 2020 http://www.jbc.org/ Downloaded from temperature for 1 to 3 h and terminated by incubation with 10 mM iodoacetate for 10-15 min at room temperature. The proteolyzed sample was then applied to a Bio-Gel A-15m column (1.5 x 41 cm) equilibrated with 100 m~ Tris, pH 7.5, at 4 "C. The vesicle-containing fractions were at the excluded volume of the column; these were pooled, dialyzed briefly to remove salt, and lyophilized.
Amino Acid Analysis and Sequencing-Automated Edman degradation was performed in a Beckman 890C Sequencer. Polybrene (2 mg) was added to the cup to minimize washout of the sample (19). The program was modeled after the standard Beckman fast 1.0 M Quadrol program, but using 0.2 M Quadrol and reduced separate benzene and ethyl acetate washes. The thiazolinones were converted to thiohydantoins with a Sequemat P-6 autoconverter in 1.5 M HCI in methanol. Phenylthiohydantoins were identified by high pressure liquid chromatography on a pBondapak CIS column in a methanol gradient (14-53%, v/v) in 0.01 M sodium acetate, pH 4.1 (20) or by amino acid analysis after hydrolysis ( 2 4 h at 110 "C in 55% HI).
Amino acid analyses were performed with a Beckman 119C amino acid analyzer. Samples were hydrolyzed for 24 h at 110 "C in sealed evacuated tubes containing either 6 N HCl or 3 M mercaptoethmesulfonic acid. Depending upon the age of the resin m d the particular amino acid analysis program, phenylalanine and tyrosine were sometimes obscured by amino sugars.
Base-catalyzed Hydrolysis of Radioactive Fractions from the Protein Sequencer, Phospholipids, and Cross-linked Peptides-Hydrolysis of the radioactive fractions from the Sequencer was performed in 0.02 N NaOH in 50% ethanol at room temperature for 1 h. The sample was then neutralized with HC1 and extracted (IO), and the extract was dried in vacuo.
Phospholipids were hydrolyzed in 0.01 N NaOH in 50% ethanol at room temperature overnight. The mixture was then neutralized and extracted with an equal volume of diethyl ether, and the organic phase was dried in vacuo.
Cross-linked glycophorin A and cross-linked proteolytic fragments rich in carbohydrate were hydrolyzed at room temperature overnight in 7.2 M urea, 0.1 N aqueous NaOH, or 0.1 N NaOH in 50% ethanol (v/ v). Cross-linked polypeptides lacking sugars were treated with 5 mM tetrabutylammonium hydroxide in dimethyl sulfoxide for 15-30 min at room temperature. The released lipid was isolated by passage through a Sephadex LH-60 column equilibrated with 8 8 % formic acid/ethanol/water (2050:30, v/v/v).
Incorporation of Phospholipid III (Fig. 1) into Red Blood Cell Ghosts. Vesicles of phospholipid 111 were prepared by drylng a chloroform solution of phospholipid 111, suspending the residue at a concentration of 0.28 mg/ml in 5 m~ sodium phosphate, pH 8.0, with vigorous Vortex mixing followed by bath sonication for 30 min. The vesicles (0.25 mg/ml) were incubated for 60 min at 37 "C with erythrocyte ghosts (1.0 mg of protein/ml) which had been prepared according to a previously described procedure (21). The sample was centrifuged for 2 min at 8000 X g. Under these conditions, the ghosts sedimented while the vesicles did not.
Reaction of Red Blood Cell Ghosts with [I4C]DCC. A solution of [I4C]DCC in ethanol (7.6 m~) was added to red blood cell ghosts suspended in 5 m~ sodium phosphate, pH 8.0, at 37 "C at a ratio of 90 nmol of ['4C]DCC/mg of protein. The protein concentration was 1-2 m g / d and the final ethanol content was 2% or less. After an overnight incubation at 37 "C, the ghosts were centrifuged as above and washed twice with 0.5 ml of ethanol.
Affinity Chromatography of Glycophorin A on Wheat Germ Agglutinin (WGA)-Sepharose-Glycophorin A was isolated from red blood cell ghosts according to Kahane et al. (22). The ghosts were solubilized with 1% SDS and after a 20-fold dilution with phosphatebuffered saline the solution was applied to a WGA-Sepharose column (0.7 x 11 cm). Glycophorin A was eluted with 1 0 0 mM N-acetylglucosamine.
Sodium Dodecyl Sulfate-Polyacrylamide Gel Electrophoresisdenaturation buffer (6% SDS, 4 M urea, 20 mM Tris, 2 mhi EDTA, 3% Samples for electrophoresis were mixed with an equal volume of mercaptoethanol, pH 6.8) and boiled for 3 min. Glycophorin A was analyzed on a 10% acrylamide gel as described (23) except that 0.5 M urea was included in the buffer. Low molecular weight fragments were analyzed with the gel system of Swank and Munkres (24). About 10 pg of SDS were added to peptides dissolved in organic solvents prior to drying. This procedure made it easier to resuspend the sample in denaturation buffer. The gels were stained with Coomassie blue and periodic acid-Schiff reagent stain (25). Lanes containing radioactive samples were sliced immediately without staining. The gel

4154
The Transmembrane Domain of Glycophorin A slices were digested with an ammonium hydroxide-hydrogen peroxide mixture and counted by liquid scintillation (26). The basic conditions reduced loss of radioactivity during the digestion, a loss presumably caused by decarboxylation of the peracid. Analytical Methods-Protein was assayed using either the ninhydrin procedure (27), or the assay using fluorescamine (28) or by the method of Lowry et al. (29). Sialic acid was determined by the method of Warren (17) and phosphate by the method of Ames and Dubin (30).

RESULTS
Incorporation of Glycophorin A into Phospholipid Vesicles-Single-walled phospholipid vesicles containing glycophorin A and DMPC or a mixture of DMPC and photolabeled phospholipids ( Fig. 1) usually at a ratio of 9 to 1 by weight were made by dialysis of a mixture of sodium cholate, glycophorin A, and DMPC as described under "Methods." A homogeneous opalescent solution resulted. The vesicles were collected by centrifugation and found to contain 60-65% of the input glycophorin A and 8045% of both phospholipids, based on sialic acid and 14C-phospholipid analyses.
The cholate dialysis procedure proved effective for the reconstitution of glycophorin A. Fig. 2A shows that if glycophorin A was added to preformed vesicles composed of DMPC and phospholipid I, the phospholipid and the protein formed separate bands on a sucrose density gradient. However, vesicles prepared by the cholate dialysis method gave a single sharp band in the density gradient tube visible to the eye which contained both the phospholipid and the protein (Fig.  2B).
One preparation of vesicles containing glycophorin A, DMPC, and phospholipid I was analyzed by freeze-fracture electron microscopy. Single lamellar vesicles were seen with a diameter of 640 f 220 A (data not shown). Intramembranous particles analogous to those seen by previous workers (31,32) were observed.
In some preparations, radioactively labeled cholate was A.

FIG. 2. Sucrose density gradient profiles of glycophorin A mixed with preformed dialyzed phospholipid vesicles (A) and of glycophorin A reconstituted into vesicles containing radioactive phospholipid I and DMPC (1:s) as under "Experimental
Procedures" (B). After centrifugation, 0.36-ml fractions were collected and analyzed for sialic acid and radioactive phospholipid.
included in the reconstitution mixture to be dialyzed. With two lots of commercial preparations of [3H]cholic acid, a substantial fraction (0.3%) of the radioactivity remained in the dialysate. However, when [3H]cholic acid pursed as described under "Methods" was used, only 0.015% of the radioactivity was retained by the vesicles, an amount corresponding to a ratio of 1 detergent molecule to 800 phospholipid molecules. This result was independent of the presence of glycophorin A in the reconstitution mixture.

Orientation of Glycophorin A in Reconstituted Vesicles-
The orientation of glycophorin A in reconstituted vesicles was determined by the addition of the membrane-impermeable enzymes neuraminidase and papain. Glycophorin A-containing vesicles were incubated with neuraminidase (1.5 units/ng of glycophorin A) at room temperature. A loss of 95 k 5% of sialic acid was observed from vesicles with the following lipid compositions: 100% DMPC; 90% DMPC and 10% phospholipid I11 (Fig. 1); or 87% DMPC and 13% phospholipid IV. The neuraminidase removed 75 +-5% of the total sialic acid from the glycophorin A in vesicles containing 90% DMPC and 10% phospholipid I. In a control, the neuraminidase removed 100% of the sialic acid from soluble glycophorin. About 75% of the sialic acid was released by papain from vesicles containing glycophorin A, 90% DMPC, and 10% phospholipid I. This proteolytic enzyme has been shown to effectively cleave glycophorin A in red blood cell ghosts (33). With both neuraminidase and papain, the reaction reached a plateau in 15 min. The remaining sialic acid was extremely stable, being uncleaved even after 3 h, demonstrating that neither neuraminidase nor papain penetrates the vesicles. The large fraction of sialic acid which was released shows that the majority of glycophorin A molecules is oriented with the NH2 terminus exposed to the outside of the vesicles.
Crcss-linking of Glycophorin A to Phospholipids on Photolysis of Vesicles-The glycophorin A vesicles described above were used to study phospholipid-protein cross-linking. Vesicles containing glycophorin A, DMPC, and phospholipid I were photolyzed at 30 "C for 4 min. The bound phospholipid (2.6%) was well separated from the free phospholipid by column chromatography on Sephadex LH-60 (Fig. 3). Little radioactivity chromatographed with glycophorin A in control experiments ( Table I) in which the vesicles were not photolyzed or in which lipid and protein were photolyzed separately or in which [14C]DMPC was used instead of one of the photolabeled lipids shown in Fig. 1.
The stability of the cross-linked protein-lipid conjugate was tested by two methods. The cross-linked protein-lipid adduct was isolated as shown in Fig. 3. The fractions (10-12, Fig. 3) containing the conjugate were pooled and then dried. The residue was redissolved and analyzed by SDS-PAGE or again passed through the Sephadex LH-60 column. In both cases, at least 8596, and usually greater than 90%, of the total radioactivity remained with the glycophorin A. Hence, by two independent methods the radioactive phospholipids associated with glycophorin A appeared to be covalently crosslinked to the protein. Glycophorin A vesicles containing the phospholipids 111-V shown in Fig. 1 were also photolyzed as Fraction number

FIG.
3. Separation of glycophorin A cross-linked to radioactive phospholipid I (Fig. 1) from monomeric and oligomeric products derived from phospholipid I on photolysis. A Sephadex LH-60 column (1.0 X 45 cm) pre-equilibrated with 88% formic acid/ethanol/water (20:50:30, v/v/v) was used. Elution was with the same solvent and the column fractions (0.8 m l ) were monitored for radioactivity (0) and for protein (0) by the ninhydrin assay. The A570 in fractions 25-38 is due to no protein, presumably Tris buffer.  (Table I). The phospholipids shown in Fig. 1B which contain a pyridyl diazirine in the polar head group reacted to a lesser extent ( Table  I).
Kinetics of Cross-linking of Phospholipids to Glycophorin A-The kinetics of lipid-lipid and lipid-protein cross-linking was similar as a function of the photolysis time (Fig. 4A). Even though cross-linking occurred mainly between phospholipid molecules, 2-3% of the photolabeled phospholipid (a substantial amount considering the ratios of the three different components in the vesicles) cross-linked to glycophorin A. The kinetic data suggested that the two cross-linkingreactions occurred by a similar mechanism, the rate-determining step being the photolytic formation of the carbene intermediate (the half-life of such intermediates may be 10-6-10-9 s or even less). However, it should be noted that, in the experiment of Fig. 4A, the aliquots from photolysis mixture were kept for about 30 min before analysis (see "Experimental Procedures"). In the next experiment, vesicles were photolyzed for 4 min and aliquots were removed at different time intervals after photolysis, with the denaturation buffer (see "Experimental Procedures"), and the extent of cross-linking was measured by SDS-PAGE. As seen in Fig. 4B, the crosslinking of phospholipid I to glycophorin A in samples kept for 1 h or more prior to the addition of denaturation buffer was 2-to %fold greater than in samples immediately mixed with denaturation buffer. Thus, at least 50-75% of the cross-linking to glycophorin A seemed to occur by a long lived intermediate and not by a carbene.
In contrast to the above results with phospholipids containing membrane-embedded photosensitive groups, no significant postphotolysis reaction was observed in experiments with vesicles prepared from phospholipid IV (data not shown).

Identification of Cross-linked Glycophorin A Peptides-
Glycophorin A cross-linked to radioactive phospholipids was isolated (Fig. 3) and cleaved using trypsin or CNBr. The insoluble tryptic fragment, T6 (residues 62-96), which has previously been isolated and characterized by Furthmayr et al. (18) was retained on a Millipore filter. In experiments with lrradlatlm tlme (sec)

FIG. 4. Time course for lipid-Fpid and lipid-protein crosslinking on photolysis of vesicles containing glycophorin A,
DMPC. and phospholipid I. A, the vesicles were photolyzed and were incubated for 30 min before analysis. The cross-linking of radioactive phospholipid I to glycophorin A was measured by chromatography on Sephadex LH-60 (Fig. 3). Lipid-lipid cross-linking was analyzed by thin layer chromatography (inset) on silica gel with Solvent 1. The phospholipid-phospholipid dimer is marked by a square in the inset. The arrow indicates the direction of solvent flow. B, shows the effect of varying the length of the incubation after irradiation. Vesicles containing radioactive phospholipid I and glycophorin A were photolyzed for 4 mi n. This longer time was given to provide for any variation in lamp intensity and to allow for decomposition of the diazo intermediate (see text). The sample was immediately mixed with denaturation buffer or incubated for different lengths of time as shown and then mixed. The extent of cross-linking was determined by SDS-PAGE. shown at the top. B, glycophorin A cross-linked to radioactive phospholipid IV (100 pg of protein, 4OOO cpm) was analyzed as described above. The low molecular weight materid (slices 15-22) is presumably due to cross-linked phospholipid released during cleavage by CNBr. Gels in both A and B were run according to Swank and Munkres (24). phospholipids I, 111, IV, and V, 96-98% of the radioactivity in the cross-linked glycophorin A after tryptic digestion was retained on the filter. In a control experiment in which trypsin was omitted, only 15-20% of the radioactivity was retained on the fdter. In some experiments, the radioactive compounds retained on the filter were eluted and electrophoresed (24). Their electrophoretic mobility was similar to that of fragment T6.
The CNBr fragments of the glycophorin A were also analyzed by gel electrophoresis (24). CNBr cleavage of glycophorin A results in four fragments previously designated by Tomita et al. (2) as follows: CB-1 (residues 9-81), CB-2 (residues 82-131), CB-3 (residues 1-8), and a partially cleaved fragment (residues 1-81). Fragment CB-3 does not appear in the SDS-PAGE analysis since it was too small to be stained. CNBr cleavage of glycophorin A cross-linked to phospholipid I gave the distribution of radioactivity shown in Fig. 5A. Thus, the bulk of radioactivity traveled with the mobility characteristic of CB-1. The radioactivity near the front (slices 15-22) was presumably nonpeptidic. Neither in this experiment nor in that performed using phospholipid 111, was any significant radioactivity found in the region of CB-2.
In experiments using phospholipids IV and V, the distribution of radioactivity in CNBr peptides was markedly different (Fig. 5B) from those described above. There was a distinct peak of radioactivity associated with CB-2.
Zsolation of Cross-linked Peptidic Fragments-In order to further characterize the site of phospholipid cross-linking, a cross-linked peptide fragment suitable for sequencing was isolated and characterized by amino acid analysis and gel electrophoresis. Glycophorin A was reconstituted into vesicles containing phospholipid I1 and DMPC (1:9 by weight). The vesicles were photolyzed and then digested with papain as under "Experimental Procedures." About 70% of the sialic acid was removed by papain, a value comparable to that observed with vesicles which had not been photolyzed. The photolyzed vesicles were then isolated on a Bio-Gel A-15m column (see "Experimental Procedures") with an 85% recovery of the radioactivity. The possibility that protein in the vesicles was loosely bound, rather than being present in a transmembranous arrangement, was tested by amino acid analysis. Phenylalanine, an amino acid present exclusively in the protected hydrophobic segment, was recovered in 75% yield.
The isolated vesicles were lyophilized and dissolved in 88% formic acid/ethanol/water (20:5030, v/v/v) and chromatographed on a Sephadex LH-60 column (Fig. 6).3 The proteincontaining fractions were labeled A, B, and C. The peaks labeled D and E had insigmikant amino acid content. Peak A stained with Coomassie blue and periodic acid-Schiff reagent stain had an apparent molecular weight lower than that of glycophorin A by 8OOO f 4000 (data not shown). Fractions B and C stained with Coomassie blue but not with PAS. The apparent molecular weights of fractions B and C were 20,000 and 12,000, respectively. Amino acid analyses of fractions A, B, and C are reported in Table 11. These were all consistent with the assignments of residues 1-105, residues 64-131, and residues 64-105 to A, B, and C, respectively. The apparent molecular weights of fractions A, B, and C indicated that they migrated as dimers during electrophoresis as did glycophorin A.
The relative molar amounts of fractions A, B, and C were 20-302 A, 55-70% B, and 0-15% C in vesicles prepared with 9 0 % DMPC and 10% phospholipid I1 and 10-20% A, 6040% B, and 0-15% C in vesicles prepared with 90% DMPC and 10% phospholipid 111. The proportions of the fractions obtained in one of the above vesicle preparations (90% DMPC + 10% phospholipid 11) was also checked by calculating the amino acid composition of the papain-cleaved vesicles from these relative mole fractions. The predicted values (Table 11, second The digestion with papain probably removed all peptides available at the external face of the vesicles since a comparable experiment with proteinase K, another nonspecific protease, gave a similar column profile (data not shown). column in parentheses) were in good agreement with the values observed.
Sequence Analysis of Phospholipid-cross-linked Peptide: The Site of Cross-linking in Glycophorin A-A prerequisite for the identification of cross-linked amino acids is the stability of the cross-linked material under the automated Edman sequencing conditions. Di-[ l-'4C]palmitoylphosphatidylcholine did not significantly degrade in heptafluorobutyric acid at 55 "C for 80 min (conditions equivalent to 20 Edman Sequencer cycles). Intact glycophorin A that had been crosslinked with phospholipid I was subjected to 10 cycles in the Sequencer; 12% of the radioactivity was released. Since proteolytic fragmentation studies discussed earlier showed no cross-linking in the NH2-terminal region, this small amount of released radioactivity was probably due to cleavage of crosslinks that were labile under sequencing conditions. Peptide B (Fig. 6) was chosen for sequence analysis because the site of phospholipid cross-linking appeared to be near its NH2 terminus and, further, it has a polar COOH terminus to retain the peptide in the Sequencer cup. The results obtained on stepwise degradation are shown in Fig. 7A. The phenylthiohydantoin amino acids identified by high performance liquid chromatography and amino acid analysis were consistent with a homogeneous NH2 terminus, Leu-64. No other phenylthiohydantoin amino acids above background were observed. The repetitive yield for sequence analysis of peptide B (residues 64-131) was 90%, a significantly higher yield than that observed (68%) for fragment T6 (residues 62-96). A large fraction (69 f 16%) of the radioactivity was released during the stepwise degradation. As seen in Fig. 7A, the radioactivity in the cross-linked phospholipid 11 was released most strongly in the benzene-ethyl acetate wash of cycle 8. Evidently, the thiazolinone corresponding to the cross-linked glutamic acid was not effectively extracted by n-chlorobutane and, instead, was extracted by ethyl acetate wash during the following cycle. Lifter et al.   6 The numbers in parentheses are the predicted values using residues 1-105,64-131, and 64-105 for A, B. and C and 0.25 A + 0.65 B + 0.10 C for the papain-cleaved vesicles. Ile 9.1 (11) 5.9 (9.2) 9.3 (10) 6.5 (9) 7.5 (8) Leu 8.0 (8) 6.6 (6.6) 8.5 (6) 6.2 (7) 6.2 (5) A, peptide B cross-linked to radioactive phospholipid I1 was generated by papain cleavage of intact vesicles and then purified by column chromatography (peak B, Fig. 6). The cross-linked peptide (92 nmol of peptide B, 24 nmol of phospholipid, 1.5 X 10' cpm) was subjected to Edman sequencing. Radioactivity at each cycle was monitored in both the chlorobutane extract and the pooled benzene-ethyl acetate extracts. The thiazolinones inthe chlorobutane extracts were determined as described under "Methods" and their identities are indicated above the corresponding extracts. B, vesicles were prepared with glycophorin A and DMPC and then reacted with [14C]DCC in the medium. The vesicles were treated with papain. Peptide B was then isolated and subjected to Edman degradation as above. A typical sequence of peptide B involved 5.3 nmol of peptide B and 9300 cpm of radioactivity. The release of radioactivity in different cycles is plotted.
Edman sequencing of glycophorin A cross-linked to phospholipids I or I11 gave similar results. Glutamic acid-70 was, therefore, accessible to the reactive intermediates of all the three phospholipids  and the location of the 14C either in the C-1 or C-2 fatty acyl chain did not significantly affect these results.

Characterization of the Major Radioactive Compound Released during the Sequencing of Cross-linked Peptide B-
The product released in cycle 7 of the Edman degradation was expected to be compound VI (Fig. 8). Benzene-ethyl acetate washes 1-13 from the Sequencer run (Fig. 7A) were dried and the residues were treated with base (see "Experimental Procedures"). Under these conditions, the linkages in compound VI would be expected to be converted to a mixture of the corresponding acid and ethyl ester. The products were analyzed by thin layer chromatography in Solvent 3 and autoradiography. Two bands ( R~0 . 7 1 and 0.75) appeared throughout the sequencing run, and under harsher hydrolysis conditions (0.1 N NaOH in 50% ethanol, overnight at room temperature) the pair of bands was converted to the slower migrating band (data not shown), suggesting that the faster-moving band was compound VII, and the slower moving band was the corresponding fatty acid, compound VIII. This assignment was confirmed by chromatographic comparison in Solvents 3 and 4 with the synthetic compounds (see "Experimental Procedures").
Release of the Radioactive Benzyl Alcohol (VIII) by Base Treatment of Cross-linked Glycophorin A-The experiments described above showed that much of the cross-linked phospholipid is attached to glutamic acid-70 in glycophorin A by an ester linkage. If so, compound VI11 (Fig. 8)   Autoradiogram of the compounds generated by basecatalyzed hydrolysis of free phospholipid, peptide A, and peptide B (Fig. 6). The thin layer chromatogram (silica gel) was developed with Solvent 4. Lane I : as a control, the material in the free phospholipid peak (Fig. 3) was treated with base. Lane 2: peptide A cross-linked with phospholipid I1 was treated with base and the hydrolysis products were separated on a Sephadex LH-60 column.
The lower molecular weight radioactive material was applied on tlc. Vesicles containing glycophorin A and phospholipid I1 were photolyzed and the cross-linked glycophorin A was separated from the free lipid on a Sephadex LH-60 column. The free lipid, cross-linked glycophorin A, and the cross-linked peptides A and B prepared as in Fig. 6 were treated with base as described under "Experimental Procedures." The release of radioactivity from glycophorin A, peptide A, and peptide B was up to 75-85%. Radioactive compounds released by base from free phospholipid, peptide A, and peptide B were analyzed by thin layer chromatography with Solvents 3 and 4. The data with the latter solvent are shown in Fig. 9. The synthetic fatty acid shown in Fig. 8 co-chromatographed with the radioactive band a, providing strong evidence that the main phospholipid-protein cross-link involved an ester linkage. Band b co-chromatographed with the fatty acid dimer described previously (6). The remaining bands were not identified.
Cross-linking Experiments Using Phospholipids IVand V (Fig. 1)-Sequencing experiments as described above were carried out with peptide B generated from glycophorin A which had been cross-linked with either phospholipid IV or V. Unsatisfactory results were obtained. Essentially all of the radioactivity was released prior to the addition of phenylisothiocyanate, indicating that the cross-links formed by these phospholipids were not stable to sequencing conditions. Therefore, the chemical nature of cross-links produced in these experiments is not clear.
Reaction which are given in Fig. 7B, showed that all of the radioactivity was released at cycle 7 and thus, glutamic acid-70 W~S the sole site of reaction of [14C]DCC with glycophorin A.
Reaction of Phospholipid 1 1 1 and ["C]DCC with Red Blood Cell Ghosts--Radioactive phospholipid 111 was introduced into the red blood cell ghosts by incubating sonicated phospholipid I11 vesicles with red blood cell ghosts at 37 "c (36)(37)(38). The efficiency of transfer of phospholipid I11 into red blood cell ghosts was 70% but was less with DMPC and phospholipids I and IV (about 20%). After the introduction of phospholipid 111, the red blood cell ghosts were photolyzed as described under "Methods" and glycophorin A was isolated by affinity chromatography on WGA-Sepharose (22). About 0.5% of the radioactivity was linked to glycophorin A.4 Fragmentation of the isolated glycophorin A with CNBr and trypsin by the procedures described above showed that greater than 95% of the radioactivity was associated with the segment containing residues 62-81 (T6 and CB-1).
Similar experhnents were carried out by incubating ["C] DCC with red blood cell ghosts (see "Methods").

DISCUSSION
Previous studies have demonstrated that glycophorin A is an integral membrane protein and it has a monotopic tripartite structure (2). One central unsolved question is the size and structure of the transmembranous segment. Our approach to this problem has been to reconstitute glycophorin A into vesicles containing photosensitive phospholipids. On irradiation, cross-linking of the phospholipids with glycophorin A would be expected to occur. The phospholipids used were such that they could interact with the protein either within the bilayer or at or near the surface of the bilayer.
For reconstitution of glycophorin A into vesicles, the cholate dialysis method was used. A similar method has recently been used for preparing unilamellar vesicles containing glycophorin A by Ong et al. (39). As described above under "Results," the vesicles were characterized in regard to their orientation both by digestion with neuraminidase and by digestion with papain. The orientation of glycophorin A was thus shown to be more than 85% right-side-out.
Photolysis of the vesicles containing glycophorin A and photoactivatable phospholipids gave a lipid-protein adduct which by several criteria contained covalent cross-links. Tryptic digestion of the cross-linked protein-phospholipid adducts showed that the hydrophobic fragment T6 (residues 62-96) was the only segment cross-linked. Edman sequence analysis was performed on the proteolytic fragment, peptide B (residues 64-131). Most of the radioactivity was released in the cycles corresponding to the thiazolinone VI (Fig. 8). Compound VI was hydrolyzed with base to yield the fatty ester/ acid (VII and VIII, Fig. 8). Their identity was confirmed by comparison with authentic samples of VI1 and VI11 prepared by chemical synthesis. The same fatty acid was also released directly from the cross-linked protein by base treatment. The sequence information and the lability to alkali showed that the carboxyl group of glutamic acid-70 was the predominant site of cross-linking.
The present work also showed that the cross-linking involved not only the insertion of the carbene intermediate formed from the diazirine into the y-carboxyl group of the glutamic acid residue but also proceeded via a long lived diazo intermediate. The photoisomerization of aryl diazirines to diazo compounds (40) and the reaction of the latter with aspartyl and glutamyl residues of proteins (41) are known. The cross-linking reactions of glycophorin A using adamantyl diazirine in a recent study (42) may also have proceeded via a similar diazo intermediate.
The cross-linking reactions using radioactive phospholipids IV and V, although they have not been investigated as thoroughly, further delineated the surface boundaries of the bilayer. Thus, following the cyanogen bromide degradation, the fragments 9-81 and 82-131 (Fig. 5B) were found labeled by the cross-linking reactions. On tryptic cleawge of the crosslinked protein, most of the radioactivity was in the fragment T6 (residues 62-96). Thus, the residues 62 and 96 must be at the bilayer surface.
Surprisingly, Glu-70 in glycophorin A was the predominant site of cross-linking both with phospholipids I and I11 which carry photosensitive fatty acids of differing chain lengths. This may be a result of two factors. First of d, there is a greater fatty acyl chain mobility toward the methyl terminus. The neutron diffraction data of Zaccai et al. (43) have provided evidence that the orientational disorder and the angular fluctuations of the segments increased at the end of the chains. The photolabel thus resides in this fluid region of the fatty acyl chain. Secondly, even though aryl diaziies are capable of labeling all amino acids, suitably positioned reactive residues such as glutamic acid or tryptophan (44) receive the bulk of the cross-linking. Segments of proteins which lack these residues might be poorly labeled even though they are in the vicinity of the photogenerated carbene.
To confirm the intramembranous location of the glutamic acid, the membrane-permeant reagent DCC was used as an independent probe. DCC is known to react specifically with glutamic or an aspartic acid residues within the membraneembedded subunits of the ATPase complexes (13,(45)(46)(47)(48).
[14C]DCC did indeed react with reconstituted glycophorin and the reaction was specific for Glu-70. Thus, the pattern of stepwise degradation and the release of radioactivity (Fig. 7B) were similar to those described above in Fig. 7A for the phospholipid I-glycophorin A cross-linked product. Corresponding experiments with red blood cell ghosts and DCC also gave similar results. Therefore, the data provide clear support for the above conclusion that Glu-70 is embedded in the membrane. Previously, Marchesi (49) made the tentative proposal that Pro-71 and Tyr-93 could be at the two boundaries of the red cell membrane. The present work shows (a) that Glu-70 is embedded in the bilayer and ( b ) that, from the labeling with phospholipids IV and V, residues 62 and 96 appear to be at the two bilayer boundaries. A model for the transmembrane organization is now proposed (Fig. 11) which is consistent with all the previous findings, summarized below, and the present results.
Measurements on lipid multilayers prepared from lipids extracted from erythrocyte ghosts (50) indicated a membrane thickness of about 45 A, and this appears to be a reasonable estimate of the bilayer thickness in red cell membranes. Circular dichroism studies of Schulte and Marchesi (51) have shown that glycophorin A is strongly a-helical between residues 62 and 96 but is largely devoid of secondary structure in Adopting an a-helical structure, Corey-Pauling-Koltum models show that Glu-70 (in the protonated form) can simultaneously hydrogen bond with Ser-69 and His-66. Also Gln-63 and His-67 can form hydrogen bonds with each other. Such hydrogen bonds would be very strong within the hydrophobic bilayer. The intramembranous a-helix of glycophorin A is amphipathic and Segrest and Jackson (55) suggested that the polar faces interact to form lateral aggregates in the membrane. Glutamic acid-72 is in the middle of the polar face and, hence, would be buried between two glycophorin A molecules. Hydrogen bonding between neighboring glycophorin A molecules through Glu-72 is a possibility.
Arginine, the most strongly charged amino acid, occupies positions 61, 96, and 97 and must be at the exterior of the bilayer. The first would delimit one side and the latter two together would define the opposite side. As stated earlier, the number of amino acids between these arginines is slightly greater than that required to traverse the membrane. This assignment is also supported by proteolysis studies in this paper.
The presence of polar or charged group, hydrogen bonds, and salt bridges is being increasingly demonstrated in membrane proteins. For example, in a number of ATPases, hydrophobic subunits contain aspartic or glutamic acid residues. A model recently proposed for another integral membrane pro-The Transmembrane Domain of Glycophorin A 4161 tein, bacteriorhodopsin, places a number of charged residues within the membrane (56). This model, as well as the present proposal for glycophorin A, emphasizes intra-and interhelical electrostatic interactions which mask the polar groups from the hydrophobic membrane.